Richard E. Plant

Classification trees: an alternative non-parametric approach for predicting species distributions.

Abstract. The use of Generalized Linear Models (GLM) in vegetation analysis has been advocated to accommodate complex species response curves. This paper investigates the potential advantages of using classification and regression trees (CART), a recursive partitioning method that is free of distributional assumptions. We used multiple logistic regression (a form of GLM) and CART to predict the distribution of three major oak species in California. We compared two types of model: polynomial logistic regression models optimized to account for non-linearity and factor interactions, and simple CART-models. Each type of model was developed using learning data sets of 2085 and 410 sample cases, and assessed on test sets containing 2016 and 3691 cases respectively. The responses of the three species to environmental gradients were varied and often non-homogeneous or context dependent. We tested the methods for predictive accuracy: CART-models performed significantly better than our polynomial logistic regression models in four of the six cases considered, and as well in the two remaining cases. CART also showed a superior ability to detect factor interactions. Insight gained from CART-models then helped develop improved parametric models. Although the probabilistic form of logistic regression results is more adapted to test theories about species responses to environmental gradients, we found that CART-models are intuitive, easy to develop and interpret, and constitute a valuable tool for modeling species distributions.

Site-specific management: the application of information technology to crop production

Site-specific management (SSM; also called, precision agriculture) is the management of agricultural crops at a spatial scale smaller than that of the whole field. Widespread farmer adoption of SSM practices is contingent on its economic advantage. Three criteria that must be satisfied in order for SSM to be justified are, (1) that significant within-field spatial variability exists in factors that influence crop yield, (2) that, causes of this variability can be identified and measured, and (3) that, the information from these measurements can be used to modify crop management practices to increase profit or decrease environmental impact. The objective of this paper is to review the state of SSM at the turn of the millennium and to offer some speculation as to its future course. The review is organized around the essential components of SSM listed above, i.e. measuring spatial variability, analyzing the data obtained from these measurements, using information gained from this analysis to effect changes in management practices, and determining whether the resulting benefits are worth the costs. The discussion section considers some potential effects of large-scale adoption of SSM, should this adoption occur.

Relationships between remotely sensed reflectance data and cotton growth and yield.

Remotely sensed electromagnetic reflectance data can provide at relatively low cost a set of detailed, spatially distributed data on plant growth and development. Vegetation indices based on algebraic combinations of different wavelength bands are especially useful in summarizing reflectance data. One of the most commonly used vegetation indices is the normalized difference vegetation index, or NDVI. The objective of this study was to determine whether measurements based on the NDVI could provide information useful for site-specific management of cotton. Aerial photographs were taken of replicated Acala cotton field experiments in California in which the treatment was water or nitrogen stress level. NDVI integrated over time showed a significant correlation with lint yield in those experiments in which there was a significant stress effect on yield. The spatiotemporal pattern of NDVI reflected stress factors and was approximately coincident with the onset of measurable water stress. NDVI tended to indicate the presence of nitrogen stress even in those cases where the stress did not result in a significant yield reduction. In a study of the correlation of NDVI with late season plant mapping indices NDVI was correlated with nodes above white flower and strongly correlated with nodes above cracked boll. An alternative vegetation index, the relative nitrogen vegetation index, was not better than NDVI as an indicator of nitrogen stress.

Bifurcation and Resonance in a Model for Bursting Nerve Cells

In this paper we consider a model for the phenomenon of bursting in nerve cells. Experimental evidence indicates that this phenomenon is due to the interaction of multiple conductances with very different kinetics, and the model incorporates this evidence. As a parameter is varied the model undergoes a transition between two oscillatory waveforms; a corresponding transition is observed experimentally. After establishing the periodicity of the subcritical oscillatory solution, the nature of the transition is studied. It is found to be a resonance bifurcation, with the solution branching at the critical point to another periodic solution of the same period. Using this result a comparison is made between the model and experimental observations. The model is found to predict and allow an interpretation of these observations.

From trickle to flood: the large-scale, cryptic invasion of California by tropical fruit flies

Since 1954, when the first tropical tephritid fruit fly was detected in California, a total of 17 species in four genera and 11 386 individuals (adults/larvae) have been detected in the state at more than 3348 locations in 330 cities. We conclude from spatial mapping analyses of historical capture patterns and modelling that, despite the 250+ emergency eradication projects that have been directed against these pests by state and federal agencies, a minimum of five and as many as nine or more tephritid species are established and widespread, including the Mediterranean, Mexican and oriental fruit flies, and possibly the peach, guava and melon fruit flies. We outline and discuss the evidence for our conclusions, with particular attention to the incremental, chronic and insidious nature of the invasion, which involves ultra-small, barely detectable populations. We finish by considering the implications of our results for invasion biology and for science-based invasion policy.

The effect of nitrogen source and crop rotation on the growth and yield of processing tomatoes

Four crop rotation and management systems were studied in 1994 and 1995 in relation to growth and yield of irrigated processing tomatoes (Lycopersicon esculentum Mill.). The four treatments were three four-year rotation systems [conventional (conv-4), low input and organic] and a two-year rotation system [conventional (conv-2)]. The four-year rotation was tomato-safflower-corn-wheat(or oats+vetch)/beans, and the two-year rotation was tomato-wheat. Purple vetch (Vicia sativa L.) was grown as a green manure cover crop preceeding tomatoes in the low input and organic systems. Nitrogen was supplied as fertilizer in the conventional systems, as vetch green manure plus fertilizer in the low input system and as vetch green manure plus turkey manure in the organic system. Tomato cv. Brigade was direct-seeded in the conventional systems and transplanted to the field in the low input and organic systems. In both years the winter cover crop was composed of a mixture of vetch and volunteer oats with N contents of 2.2% in 1994 and 2.7% (low input) or 1.8% (organic) in 1995. In 1994 yields were higher in conventionally grown tomatoes because a virus in the nursery infected the transplants used in the low input and organic systems. In 1995 tomatoes grown with the low input and conv-4 systems had similar yields, which were higher than those of tomatoes grown with the conv-2 and organic systems. N uptake by the crop was greater than 200 kg N ha−1 for high yield (> 75 t ha−1) and uptake rates of 3 to 6 kg N ha−1 day−1 during the period of maximum uptake were observed. The lower yield with the organic system in 1995 was caused by a N deficiency. The main effect of the N deficiency was a reduced leaf area index and not a reduction of net assimilation rate (NAR) or radiation use efficiency (RUE). Nitrogen deficiency was related to low concentration of inorganic N in the soil and slow release of N from the cover crop + manure. A high proportion of N from the green manure but only a low proportion of N from the manure was mineralized during the crop season. In the conventional systems, the estimated mineralized N from the soil organic matter during the crop season was around 85 kg ha−1. A hyperbolic relationship between N content and total dry weight of aboveground biomass was observed in procesing tomatoes with adequate N nutrition. Lower yields with the conv-2 than with the conv-4 system were due to higher incidence of diseases in the two year rotation which reduced the NAR and the RUE. Residual N in the soil in Oct. (two months after the incorporation of crop residues) ranged between 90 and 170 kg N ha−1 in the 0–90 cm profile.

An integrated expert decision support system for agricultural management

Abstract This paper describes the CALEX system, a microcomputer based integrated expert decision support system for agricultural management. The program, which is implemented in conjunction with domain specific modules, consists of three separate subprograms: an executive, a scheduler, and an expert system shell. The executive serves as the primary interface to the user, to models, and to the disk. The scheduler generates a sequence of management activities by repeatedly activating the expert system. The expert system makes the actual management decisions. The CALEX package is domain independent and can be used with any commodity. Initial development of the program has focused on the development of a package of modules for California cotton and another package for peaches. This paper describes the architecture of the basic CALEX program. The features of the CALEX package are illustrated with brief descriptions of some of the commodity specific implementations. Full descriptions of commodity specific implementations will be given elsewhere.

A FitzHugh Differential-Difference Equation Modeling Recurrent Neural Feedback

A very simple model based on the FitzHugh equations is developed to simulate the phenomenon of recurrent neural feedback. This phenomenon, which is ubiquitous in the vertebrate nervous system, occurs when a neuron excites a second neuron which in turn excites or inhibits the first neuron. Since the excitation or inhibition occurs only after conduction and synaptic delays, the model involves a system of differential-difference equations. Conditions for the existence of a Hopf bifurcation are derived, and formulas for the stability of the bifurcation are given. Some numerical results for large amplitude solutions are presented. A discussion of the applicability of the model is given.

Application of Epic Model to Nitrogen Cycling in Irrigated Processing Tomatoes Under Different Management Systems

Abstract Vegetable crops such as processing tomatoes (Lycopersicon esculentum Mill.) are usually complex in terms of nitrogen (N) dynamics because of the large amounts absorbed by the crop, the short growing season and the use of irrigation. Complexity increases when N is supplied from an organic source. A crop simulation model could be very useful to improve N management in this crop. Processing tomatoes were grown on raised beds and furrow irrigated in 1994 and 1995 in the Sacramento Valley of California. Fertilizer N and/or purple vetch (Vicia sativa L.) as green manure and composted turkey manure were used as sources of N. The Erosion Productivity Impact Calculator (EPIC) model was calibrated with 1994 data and validated with 1995 data. Plant growth was accurately simulated in the conventional systems that used fertilizer N and in the low input system that used fertilizer N plus vetch. The model accurately simulated above-ground biomass in a system that used vetch and no synthetic fertilizer N, but it over-predicted Leaf Area Index (LAI). Nitrogen deficiency was observed in the plants in this system. The model simulated nitrogen deficiency mainly as a reduction in biomass production but in the real world the reduction of leaf area was the first effect of nitrogen deficiency in the vegetative phase. Yields were accurately predicted except when diseases affected plant growth. A simple reduction factor of nitrate movement in the bed adequately addressed the movement of nitrate. In general, the model accurately predicted the evolution of inorganic nitrogen in different soil layers during the crop season. However, simulated inorganic N in the upper 15 cm was underestimated in the last part of the crop season and consequently N uptake at harvest was slightly over-predicted in some cases. Nitrogen distribution and access of the roots to inorganic nitrogen are discussed as causes of this discrepancy between model simulated and observed values.

Object-Based Image Classification of Summer Crops with Machine Learning Methods

The strategic management of agricultural lands involves crop field monitoring each year. Crop discrimination via remote sensing is a complex task, especially if different crops have a similar spectral response and cropping pattern. In such cases, crop identification could be improved by combining object-based image analysis and advanced machine learning methods. In this investigation, we evaluated the C4.5 decision tree, logistic regression (LR), support vector machine (SVM) and multilayer perceptron (MLP) neural network methods, both as single classifiers and combined in a hierarchical classification, for the mapping of nine major summer crops (both woody and herbaceous) from ASTER satellite images captured in two different dates. Each method was built with different combinations of spectral and textural features obtained after the segmentation of the remote images in an object-based framework. As single classifiers, MLP and SVM obtained maximum overall accuracy of 88%, slightly higher than LR (86%) and notably higher than C4.5 (79%). The SVM+SVM classifier (best method) improved these results to 89%. In most cases, the hierarchical classifiers considerably increased the accuracy of the most poorly classified class (minimum sensitivity). The SVM+SVM method offered a significant improvement in classification accuracy for all of the studied crops compared to the conventional decision tree classifier, ranging between 4% for safflower and 29% for corn, which suggests the application of object-based image analysis and advanced machine learning methods in complex crop classification tasks.